Hi Wilko, That's great - your version of my configuration works immediately without the read-only problem I was having (and no other changes needed.) The only tweak I needed was adding xrootd.allow host disk042.gla.scotgrid.ac.uk cmsd.allow host disk042.gla.scotgrid.ac.uk So, now I'm trying something a little more advanced: I replaced the all.export with oss.cache public /gridstore0 xa oss.cache public /gridstore1 xa oss.cache public /gridstore2 xa all.export /disk r/w and added an xrootd.allow host disk*.gla.scotgrid.ac.uk cmsd.allow host disk*.gla.scotgrid.ac.uk in place of the specific disk042 I then stuck identically configured xrootd + cmsd services on other disk servers in this range, and could see them connecting to svr025 after bringing it's services back up. I think that this should allow me to do xrdcp somefile root://svr025.gla.scotgrid.ac.uk//disk/file1 and have the file1 actually be stored on one of the gridstore filesystems. What actually happens is that I get: No servers are available to write the file. Which looks like something is confused somewhere? Do I need to explictly tell xroot that it needs to use the cache as the storage location for /disk related transfers? Sam 2009/5/6 Wilko Kroeger <[log in to unmask]>: > > Hello Sam > > Just a few comments. > As you found out the example config files on the slac web page are quite old > and out dated. I will take them down and replace them with a newer version. > The configuration options changed a bit they are easier now. > Also if possible you should use the cmsd instead of the olbd, > I modified your config files and made one out of it that will work with both > your data server and redirector. > > # config file for both dataserver and redirector > > all.manager svr025.gla.scotgrid.ac.uk 2526 > all.export /gridstore4 r/w > if svr025.gla.scotgrid.ac.uk > all.role manager > else > all.role server > fi > > xrootd.fslib /root/20090202-1402/lib/libXrdOfs.so > xrootd.allow host svr025.gla.scotgrid.ac.uk > > xrd.protocol xrootd * > xrd.port 2525 > > cms.allow host svr025.gla.scotgrid.ac.uk > cms.trace redirect > > > In order to start the xrootd and cms you can run: > /root/20090202-1402/bin/xrootd -c <configfile> -l <logfile> > /root/20090202-1402/bin/cmsd -c <configfile> -l <logfile> > > If there is already a xrootd running on one of these machine you have to use > the name option (-n) for the xrootd/cmsd, e.g.: > /root/20090202-1402/bin/xrootd -n mytest -c <configfile> -l <logfile> > /root/20090202-1402/bin/cmsd -n mytest -c <configfile> -l <logfile> > > You could also use the StartXRD, StartCMS scripts which are in the etc dir > of the xrootd release but for this you have to configure the StartXRD.cf > script. > > I hope this helps. Please let me know if you have more questions. > > Cheers, > Wilko > > > On Wed, 6 May 2009, Fabrizio Furano wrote: > >> Hi Sam, >> >> so, if your preference is for the cern all-in-one bundle, then all the >> instructions are in the cern url which I sent to you. >> >> Changing the ports is trivial and does not require low level tweaking, >> since the values are in the meta-configuration file system.cnf. Just put >> reasonable values following the criteria which are there. >> >> My advice is to setup/run tests in a machine which is not in production, >> even a linux laptop is 100% ok. Just for convenience and ease of operation. >> >> Just a question: what is your test setup supposed to be for? >> >> Fabrizio >> >> Sam Skipsey wrote: >>> >>> Hi Fabrizio, >>> >>> So, I tried setting oss.localroot (and r/w) , but the data server >>> still insists that it doesn't have write access to the filesystem and >>> so sets itself to read only. >>> Since it is being run as dpmmgr user, and the filesystem is owned by >>> that user and is mode 0700, I'm at a bit of a loss to explain this. >>> >>> Thank you for the CERN link - I've been using the packages from the >>> slac link, so at least the version of xrootd I have is reasonably up >>> to date. Unfortunately, I doubt that I can directly use the all-in-one >>> version without modifying it anyway, since I need to run the services >>> on non-standard ports - I'm trying to run this initial test xrootd >>> service on a head node which already hosts a dpm server (and >>> dpm-xrootd plugins), so I can't use the default ports. >>> (The style of the configuration files is based on the Example files >>> provided on the SLAC website. Perhaps the examples need a bit of >>> updating ;) ) >>> >>> Sam >>> >>> 2009/5/6 Fabrizio Furano <[log in to unmask]>: >>>> >>>> Hi Sam, >>>> >>>> you should specify r/w in the export directive of the data server, >>>> writing >>>> this: >>>> >>>> xrootd.export /gridstore4 r/w >>>> >>>> Then please verify that you are specifying the correct localroot >>>> option. >>>> >>>> Moreover, the config files you wrote look really obsolete, and probably >>>> valid for very old versions. My suggestion is to get a newer package and >>>> refer to the online documentation to configure it. Things are much >>>> simpler >>>> now, in the sense that fewer directives are needed. >>>> >>>> Depending on your preferences, there are two flavors of the bundle >>>> which >>>> you can get: the vanilla one and the 'automatic all-in-one' one. >>>> Respectively, you can find the first here: >>>> >>>> http://xrootd.slac.stanford.edu >>>> (or you can take the tarball from the last ROOT version) >>>> >>>> and the second here: >>>> >>>> http://savannah.cern.ch/projects/xrootd >>>> >>>> Fabrizio >>>> >>>> >>>> Sam Skipsey wrote: >>>>> >>>>> Hello, all: >>>>> >>>>> I may be being stupid here, but: >>>>> >>>>> I'm in the initial phase of messing around with xrootd here at >>>>> Glasgow, and I can't seem to get my diskserver to add itself as >>>>> writeable. >>>>> >>>>> The arrangement is: >>>>> svr025.gla.scotgrid.ac.uk running as a redirection server (xrootd on >>>>> port 2525 and olbd on 2526) >>>>> disk042.gla.scotgrid.ac.uk (for example) running as a data server >>>>> (same services on same ports) >>>>> >>>>> The config for the two is: >>>>> >>>>> redirector: >>>>> # >>>>> # redirectserver.cf >>>>> # >>>>> # xrootd >>>>> xrootd.port 2525 >>>>> xrootd.fslib /root/20090202-1402/lib/libXrdOfs.so >>>>> xrootd.export /gridstore4 >>>>> #oss.readonly >>>>> odc.manager svr025.gla.scotgrid.ac.uk 2526 >>>>> odc.trace redirect >>>>> # olbd >>>>> olb.port 2526 >>>>> olb.allow host disk042.gla.scotgrid.ac.uk >>>>> olb.allow host svr025.gla.scotgrid.ac.uk >>>>> ofs.redirect remote svr025.gla.scotgrid.ac.uk >>>>> xrootd.allow host svr025.gla.scotgrid.ac.uk >>>>> >>>>> >>>>> dataserver: >>>>> # >>>>> # dataserver.cf >>>>> # >>>>> # xrootd >>>>> xrootd.port 2525 >>>>> xrootd.fslib /root/20090202-1402/lib/libXrdOfs.so >>>>> xrootd.export /gridstore4 >>>>> #oss.readonly >>>>> odc.manager svr025.gla.scotgrid.ac.uk 2526 >>>>> # olbd >>>>> olb.port 2526 >>>>> olb.subscribe svr025.gla.scotgrid.ac.uk 2526 >>>>> ofs.redirect remote svr025.gla.scotgrid.ac.uk >>>>> oft.redirect target >>>>> >>>>> >>>>> However, when disk042 connects to the redirection server, I see: >>>>> >>>>> 090506 14:19:21 31690 olb_Manager: server disk042 defaulted r / >>>>> >>>>> and on the disk: >>>>> 090506 14:18:01 001 olb_Meter: Warning! No writable filesystems found; >>>>> write access and staging prohibited. >>>>> >>>>> and, indeed, xrdcp to disk042 fails (but copying from it works >>>>> perfectly). >>>>> The exported filesystem (/gridstore4) is writable by the user owning >>>>> the xrootd processes, however (indeed, it is owned by the same user, >>>>> with mode 770). >>>>> >>>>> This is probably something stupid, but... any help appreciated. >>>>> >>>>> Sam Skipsey >>>>> Glasgow >> >